The Global Positioning System (GPS) long-baseline set up has been widely employed to generate high-accuracy positioning, timing and atmospheric information. Bernese GPS software adopts two appropriate strategies for long-baseline Integer Ambiguity Resolution (IAR): Quasi Ionosphere-Free (QIF) and Wide-lane/Narrow-lane (WN). With the goal of reasonably shortening the time required for long-baseline IAR, we propose the Precise Point Positioning (PPP) method for estimating, on a per receiver basis, the Zenith Tropospheric Delays (ZTDs) and the Slant Ionospheric Delays (SIDs) from zero-differenced, uncombined GPS observables. We then reformulate these PPP-derived ZTDs and SIDs into two types of atmospheric constraints with proper uncertainties that could be readily assimilated into the process of IAR with the QIF. Our numerical tests based on five independent long-baselines (>1,000 kilometres) suggest that the empirical precision of PPP-derived ZTDs (SIDs) is always better than 2 (10) centimetres. The modified QIF would be able to correctly resolve at least 98% and 88% of the wide- and narrow-lane ambiguities for all the long-baselines relying on the very simple integer rounding method. However, under the same condition, the WN can only get the correct integers of 76·6% wide-lane ambiguities and 55·2% narrow-lane ones.